meilisearch.models package

Submodules

meilisearch.models.document module

class meilisearch.models.document.Document(doc: dict[str, Any])[source]

Bases: object

class meilisearch.models.document.DocumentsResults(resp: dict[str, Any])[source]

Bases: object

class meilisearch.models.document.FieldsResults(resp: dict[str, Any])[source]

Bases: object

Response object for get_fields containing pagination metadata and field list.

meilisearch.models.embedders module

class meilisearch.models.embedders.CompositeEmbedder(*args: Any, **kwargs: Any)[source]

Bases: CamelBase

Composite embedder configuration.

Parameters:

  • source (str) – The embedder source, must be “composite”

  • indexing_embedder (Union[) – OpenAiEmbedder, HuggingFaceEmbedder, OllamaEmbedder, RestEmbedder, UserProvidedEmbedder,

  • ]

  • search_embedder (Union[) – OpenAiEmbedder, HuggingFaceEmbedder, OllamaEmbedder, RestEmbedder, UserProvidedEmbedder,

  • ]

indexing_embedder: OpenAiEmbedder | HuggingFaceEmbedder | OllamaEmbedder | RestEmbedder | UserProvidedEmbedder
search_embedder: OpenAiEmbedder | HuggingFaceEmbedder | OllamaEmbedder | RestEmbedder | UserProvidedEmbedder
source: str = 'composite'
class meilisearch.models.embedders.Distribution(*args: Any, **kwargs: Any)[source]

Bases: CamelBase

Distribution settings for embedders.

Parameters:

  • mean (float) – Mean value between 0 and 1

  • sigma (float) – Sigma value between 0 and 1

mean: float
sigma: float
class meilisearch.models.embedders.Embedders(*args: Any, **kwargs: Any)[source]

Bases: CamelBase

Container for embedder configurations.

Parameters:

embedders (Dict[str, Union[OpenAiEmbedder, HuggingFaceEmbedder, OllamaEmbedder, RestEmbedder, UserProvidedEmbedder]]) – Dictionary of embedder configurations, where keys are embedder names

embedders: dict[str, EmbedderType]
class meilisearch.models.embedders.HuggingFaceEmbedder(*args: Any, **kwargs: Any)[source]

Bases: CamelBase

HuggingFace embedder configuration.

Parameters:

  • source (str) – The embedder source, must be “huggingFace”

  • url (Optional[str]) – The URL Meilisearch contacts when querying the embedder

  • model (Optional[str]) – The model your embedder uses when generating vectors (defaults to BAAI/bge-base-en-v1.5)

  • dimensions (Optional[int]) – Number of dimensions in the chosen model

  • revision (Optional[str]) – Model revision hash

  • document_template (Optional[str]) – Template defining the data Meilisearch sends to the embedder

  • document_template_max_bytes (Optional[int]) – Maximum allowed size of rendered document template (defaults to 400)

  • distribution (Optional[Distribution]) – Describes the natural distribution of search results

  • binary_quantized (Optional[bool]) – Once set to true, irreversibly converts all vector dimensions to 1-bit values

  • pooling (Optional[PoolingType]) – Configures how individual tokens are merged into a single embedding

binary_quantized: bool | None = None
dimensions: int | None = None
distribution: Distribution | None = None
document_template: str | None = None
document_template_max_bytes: int | None = None
model: str | None = None
pooling: PoolingType | None = 'useModel'
revision: str | None = None
source: str = 'huggingFace'
url: str | None = None
class meilisearch.models.embedders.OllamaEmbedder(*args: Any, **kwargs: Any)[source]

Bases: CamelBase

Ollama embedder configuration.

Parameters:

  • source (str) – The embedder source, must be “ollama”

  • url (Optional[str]) – The URL Meilisearch contacts when querying the embedder (defaults to http://localhost:11434/api/embeddings)

  • api_key (Optional[str]) – Authentication token Meilisearch should send with each request to the embedder

  • model (Optional[str]) – The model your embedder uses when generating vectors

  • dimensions (Optional[int]) – Number of dimensions in the chosen model

  • document_template (Optional[str]) – Template defining the data Meilisearch sends to the embedder

  • document_template_max_bytes (Optional[int]) – Maximum allowed size of rendered document template (defaults to 400)

  • distribution (Optional[Distribution]) – Describes the natural distribution of search results

  • binary_quantized (Optional[bool]) – Once set to true, irreversibly converts all vector dimensions to 1-bit values

api_key: str | None = None
binary_quantized: bool | None = None
dimensions: int | None = None
distribution: Distribution | None = None
document_template: str | None = None
document_template_max_bytes: int | None = None
model: str | None = None
source: str = 'ollama'
url: str | None = None
class meilisearch.models.embedders.OpenAiEmbedder(*args: Any, **kwargs: Any)[source]

Bases: CamelBase

OpenAI embedder configuration.

Parameters:

  • source (str) – The embedder source, must be “openAi”

  • url (Optional[str]) – The URL Meilisearch contacts when querying the embedder

  • api_key (Optional[str]) – Authentication token Meilisearch should send with each request to the embedder

  • model (Optional[str]) – The model your embedder uses when generating vectors (defaults to text-embedding-3-small)

  • dimensions (Optional[int]) – Number of dimensions in the chosen model

  • document_template (Optional[str]) – Template defining the data Meilisearch sends to the embedder

  • document_template_max_bytes (Optional[int]) – Maximum allowed size of rendered document template (defaults to 400)

  • distribution (Optional[Distribution]) – Describes the natural distribution of search results

  • binary_quantized (Optional[bool]) – Once set to true, irreversibly converts all vector dimensions to 1-bit values

api_key: str | None = None
binary_quantized: bool | None = None
dimensions: int | None = None
distribution: Distribution | None = None
document_template: str | None = None
document_template_max_bytes: int | None = None
model: str | None = None
source: str = 'openAi'
url: str | None = None
class meilisearch.models.embedders.PoolingType(value)[source]

Bases: str, Enum

Pooling strategies for HuggingFaceEmbedder.

USE_MODEL

Use the model’s default pooling strategy.

Type:

str

FORCE_MEAN

Force mean pooling over the token embeddings.

Type:

str

FORCE_CLS

Use the [CLS] token embedding as the sentence representation.

Type:

str

FORCE_CLS = 'forceCls'
FORCE_MEAN = 'forceMean'
USE_MODEL = 'useModel'
class meilisearch.models.embedders.RestEmbedder(*args: Any, **kwargs: Any)[source]

Bases: CamelBase

REST API embedder configuration.

Parameters:

  • source (str) – The embedder source, must be “rest”

  • url (Optional[str]) – The URL Meilisearch contacts when querying the embedder

  • api_key (Optional[str]) – Authentication token Meilisearch should send with each request to the embedder

  • dimensions (Optional[int]) – Number of dimensions in the embeddings

  • document_template (Optional[str]) – Template defining the data Meilisearch sends to the embedder

  • document_template_max_bytes (Optional[int]) – Maximum allowed size of rendered document template (defaults to 400)

  • indexing_fragments (Optional[Dict[str, Dict[str, Any]]]) – Defines how to fragment documents for indexing (multi-modal search). Fragments can contain complex nested structures (e.g., lists of objects).

  • search_fragments (Optional[Dict[str, Dict[str, Any]]]) – Defines how to fragment search queries (multi-modal search). Fragments can contain complex nested structures (e.g., lists of objects).

  • request (Dict[str, Any]) – A JSON value representing the request Meilisearch makes to the remote embedder

  • response (Dict[str, Any]) – A JSON value representing the request Meilisearch expects from the remote embedder

  • headers (Optional[Dict[str, str]]) – Custom headers to send with the request

  • distribution (Optional[Distribution]) – Describes the natural distribution of search results

  • binary_quantized (Optional[bool]) – Once set to true, irreversibly converts all vector dimensions to 1-bit values

api_key: str | None = None
binary_quantized: bool | None = None
dimensions: int | None = None
distribution: Distribution | None = None
document_template: str | None = None
document_template_max_bytes: int | None = None
headers: dict[str, str] | None = None
indexing_fragments: dict[str, dict[str, Any]] | None = None
request: dict[str, Any]
response: dict[str, Any]
search_fragments: dict[str, dict[str, Any]] | None = None
source: str = 'rest'
url: str | None = None
class meilisearch.models.embedders.UserProvidedEmbedder(*args: Any, **kwargs: Any)[source]

Bases: CamelBase

User-provided embedder configuration.

Parameters:

  • source (str) – The embedder source, must be “userProvided”

  • dimensions (int) – Number of dimensions in the embeddings

  • distribution (Optional[Distribution]) – Describes the natural distribution of search results

  • binary_quantized (Optional[bool]) – Once set to true, irreversibly converts all vector dimensions to 1-bit values

binary_quantized: bool | None = None
dimensions: int
distribution: Distribution | None = None
source: str = 'userProvided'

meilisearch.models.index module

meilisearch.models.key module

meilisearch.models.task module

meilisearch.models.webhook module

Module contents